Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

2016 Loughborough Antennas & Propagation Conference (LAPC)

Subarrays in linear array configurations,


an effective instrument for scan loss compensation

Fannush Shofi Akbar†, Leonardus P. Ligthart‡ , Ioan E. Lager‡ , and Gamantyo Hendrantoro†

Institut Teknologi Sepuluh Nopember, Department of Electrical Engineering, Surabaya 60111, Indonesia,
Email: fannush.akbar13@mhs.ee.its.ac.id

Delft University of Technology, Faculty of Electrical Engineering, Mathematics and Computer Science,
Mekelweg 4, 2628 CD Delft, the Netherlands

Abstract—A novel approach is proposed for compensating time, maximising the number of empty spaces large enough
the scan loss in wide-angular scanning array antennas due to for accommodating the optimised subarrays. The radiation
the pattern falloff of physical radiators. The method is based performance of these configurations are discussed in detail.
on replacing elements in (linear) arrays with subarrays with
enhanced radiation intensity at large angles. This strategy is The encouraging obtained results allow conjecturing that
best applicable in the case of sparse arrays that intrinsically this strategy, when applied to planar arrays, will lead to
provide the space for accommodating subarrays. An optimised an important scan-loss reduction in planar arrays. Such an
linear subarray, with 5 dB radiation intensity enhancement at 60◦ improvement is conducive to substantially enhancing the per-
is designed. This subarray is then incorporated in linear array formance of wide-angular scanning radar array antennas.
antennas with (approximately) 50% thinning factors. An up to
1.2 dB improvement of the maximum directivity at 60◦ beam
scanning is obtained. The obtained results allow conjecturing
II. D ESIGN PHILOSOPHY
that a similar strategy, when applied to planar arrays, will lead A. Prerequisites
to an even more significant scan-loss reduction, this yielding a
substantial performance enhancement for wide-angular scanning Let the case of the linear array in Fig. 1.a, consisting
radar array antennas. of N radiators located along the Ox axis at r n = xn x̂,
m = 1, 2, . . . , N and x1 = 0, with x̂ being the unit
I. I NTRODUCTION vector along the Ox axis. As habitually with linear arrays [4,
Chapter 6], the radiation pattern is considered with respect to
Modern radar applications place high demands on the the angle ϑ ∈ [−π/2, π/2], with ϑ measuring the tilting from
electronic beam scanning of the pertaining antenna systems. In the x = 0 plane and assuming positive values in the half-
particular, circular beam scanning via facetted array antennas space x > 0. The excitation is time-harmonic, with frequency
often requires extremely wide beam scanning, with a beam f . The excitations at all feeding ports have unit amplitude and
scanning from −60◦ up to 60◦ with respect to the facet’s are in phase for broadside radiation and have progressive phase
broadside being often specified. Achieving such performance shifts for beam scanning. The array radiates in free space, the
requires dealing with important design and technological dif- wavespeed being c0 and the corresponding wavelength being
ficulties. One such challenge follows from the intrinsic falloff λ = c0 /f . The radiators are taken to radiate in the upper half-
of the radiation pattern of physical radiators. For example, a space z > 0, only. The array’s directivity [4, Section 2.2] can
cos(ϑ)-type radiator exhibits a 6 dB loss at 60◦ . The perfor- then be expressed as
mance degradation is even more accentuated in the case of
dense, planar arrays that are known to have lower directivity κ 2 [E(ϑ) · E ∗ (ϑ)]
D(ϑ) = (1)
characteristics with scanning. This situation is inacceptable for 
π/2

many radar applications and remedies are stringently needed. [E(ϑ) · E ∗ (ϑ)] dϑ
−π/2
This paper introduces a novel approach to tackling this
where E(ϑ) is the electric field strength on the far-field sphere
problem, by using subarrays with enhanced radiation intensity
of radius R  λ/2π and ‘*’ denotes complex conjugation.
at large angles. Such subarrays could be appended at the edges
The parameter κ accounts for the one-directional radiation and
of dense arrays. However, a much more effective approach is
takes the values 1 for half-space normalisation and 2 for full-
to opportunely make use of the free space inside sparse arrays,
space normalisation.
this allowing accommodating a substantially larger number of
subarrays on, essentially, the same overall real estate, with the
B. Subarrays as a means to compensate the scan loss
ensuing improved scan-loss compensation.
As stated in the Introduction, the elementary radiation
For clarity, this paper is restricted to the investigation of
pattern directly impacts on the radiated field, this effect being
linear arrays. Firstly a linear subarray with a 5 dB radiation
highly relevant in the case of wide-scanning arrays. For exam-
intensity enhancement at 60◦ is designed. Two prototype linear
ple, a cos(ϑ)-type radiator exhibits an intrinsic loss of 6 dB at
arrays are then designed by means of the dynamic program-
60◦ , this being inacceptable for many radar applications.
ming (DynP) [1] and the Cyclic Difference Sets (CDS) [2],
[3] placement strategies. The array design aims at providing An expedient way for precluding this loss is by resorting
an (approximately) 50% thinning factor while, at the same to subarrays with a convenient radiation pattern that, ideally,

978-1-5090-0783-7/16/$31.00 ©2016 IEEE


2016 Loughborough Antennas & Propagation Conference (LAPC)

z
I =Phase shifter The configuration is shown in Fig. 2, the primary element
ϑ
being located at xn and the parasitic elements being spaced
x at λ/2. The subarray radiators are fed via a power divider
O x1 x2 xn xN −1 xN
attenuator yielding the elementary amplitudes An , An− and
Feeding
network An+ , respectively, with An , An− , An+ ∈ R for ensuring an
a equal phase feeding.
z
5.5 -0.14
-0.14 dBi
dBi 4.87
4.87 dBi
dBi
ϑ 5
4.5
x 4
O x1 x2 xn xN −1 xN
3.5
Feeding

|D(ϑ)| [dBi]
network 3
2.5
b 2

Fig. 1. Investigated configurations. (a) Linear array consisting of N radiators 1.5

located along the Ox-axis; each element is fed via an individual phase shifter. 1
(b) Same array, with some of the elements being replaced by subarrays. 0.5
0

−90 −60 −30 0 30 60 90


ϑ [deg]
should fully compensate the elementary pattern’s falloff. This
idea is illustrated in Fig. 1.b in which some of the elements in Fig. 3. Directivity pattern (κ = 1) of the optimised X3-sa.
the original linear array in Fig. 1.a are replaced by subarrays
consisting of a primary element, collocated with the original The quantities An , An− and An+ were optimised via a grid
element, enclosed between two parasitic elements. The intro- search strategy for maximising the radiation at the intended
duction of the subarray should not perturb the (presumably, end-of-coverage angle of 60◦ . The procedure yielded An =
optimised) array factor (AF) of the original array. To that end, 1 and An− = An+ = −0.145, the corresponding radiation
the phase centre of the (meta-)element at the relevant location pattern being shown in Fig. 3. The radiation intensity is slightly
may not shift. If all 3 elements in the subarray are fed with diminished at broadside, namely D(0◦ ) = −0.14 dBi, but is
the same phase and the subarray is symmetric, elementary significantly enforced at 60◦ , with D(60◦ ) = 4.87 dBi. While
results in Fourier transform theory [5, p. 13] directly state the latter value does not fully compensate the 6 dBi loss of
that the subarray’s AF is real and symmetric. Consequently, a cos(ϑ)-type pattern, this design offers a good compromise
the AF’s phase is constant on the far-field sphere, as in the between performance and the subarray physical size.
case with isotropic radiators. Such subarrays then comply to
the requirement of not shifting the phase centre and will be IV. P ERFORMANCE ANALYSIS
henceforth consistently used.
The scan loss compensation strategy advocated in Sec-
Evidently, the linear array’s architecture must offer suf- tion II-B is now demonstrated for the case of two linear
ficient room around the elements to be substituted for ac- arrays that are deemed interesting for practical applications.
commodating the intended subarrays. For uniform arrays, this The performance of the original linear arrays will be firstly
will imply a two wide inter-element spacing that is likely discussed and the improvement as a result of introducing the
to result in the onset of grating lobes (GL-s). Resorting X3-sa-s will be then examined. This analysis will focus on 3
to non-uniform arrays seems unavoidable. Several effective, scanning angles, namely ϑs = 0◦ , ϑs = 30◦ and ϑs = 60◦ .
well tested strategies for designing sparse arrays are readily
available. In this paper, we focus on the DynP and CDS A. Thinned, linear arrays
placements, with possible alternatives being found in [6]–[9].
In this paper, we analyse thinned, sparse, linear arrays
using a λ/2-spaced generating lattice. The envisaged thinning
III. 3- ELEMENTS SUBARRAY DESIGN factor is 50%. Two thinning techniques that are known to
preclude the onset of GL-s are employed. In both cases,
ϑ the design aims at obtaining comparable array lengths. The
key objective is enhancing the number of isolated elements,
namely elements with no immediate neighbours, these being
the evident candidates for being replaced by X3-sa-s.
xn −λ/2 xn xn +λ/2
The first linear array was designed via the DynP [1]
An− An An+ placement strategy. The configuration is shown in Fig. 4.a. It
replicates the one discussed in [10], where details on the design
method can also be found. The resulting array has 30 elements
and its length is 29.5λ. The second array was obtained via
Fig. 2. Generic X3-sa consisting of a main radiator and two λ/2-spaced the CDS placement [2], [3], by complementing the placement
parasitic elements, fed via a power divider attenuator. following from the by 35 cyclically shifted, {59, 29, 14} CDS
available in [11]. The resulting array is shown in Fig. 4.b. It
A symmetric array of 3 elements, hereafter referred to has the same number of elements as the previous one, but its
as X3-sa, was designed for being employed as subarray. length is marginally smaller, namely 29λ.

978-1-5090-0783-7/16/$31.00 ©2016 IEEE


2016 Loughborough Antennas & Propagation Conference (LAPC)

z
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 30
TABLE I. K EY ARRAY PERFORMANCE METRICS FOR THE ARRAY
x FACTORS IN F IG . 5.
O
D(ϑ)max Beamwidth First SLL Peak SL
29.5 λ
a DynP placement
z
ϑs = 0◦ 14.8 dBi 2.6◦ -16.3 dB -14.6 dB
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 30
x ϑs = 30◦ 14.8 dBi 3◦ -16.3 dB -14.6 dB
O ◦
ϑs = 60 14.8 dBi 5.2◦ -16.3 dB -14.6 dB
29 λ CDS placement
b ϑs = 0◦ 14.8 dBi 1.8◦ -13.3 dB -12.9 dB
ϑs = 30◦ 14.8 dBi 2.1◦ -13.3 dB -12.9 dB
Fig. 4. Thinned arrays, starting from a λ/2-spaced, uniform, linear array. ϑs = 60◦ 14.8 dBi 3.7◦ -13.3 dB -12.9 dB
(a) Dynamic programming placement; (b) CDS placement based on a by 35
cyclically shifted {59, 29, 14}-CDS array – complemented placement.
B. Effect of the use of X3-sa-s
The X3-sa discussed in Section III is now used in the
16 14.8
14.8 dBi
dBi 14.8
14.8 dBi
dBi 14.8
14.8 dBi
dBi
ϑsc = 0◦
14
ϑsc = 30◦
12
ϑsc = 60◦
array configurations depicted in Fig. 4. It is clear that the DynP
10
8
design allows for using a maximum of 4 X3-sa-s, whereas the
CDS design can accommodate up to 7 X3-sa-s (see Fig. 6).
|D(ϑ)| [dBi]

6
4
z
2
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 30
0 x
−2 O
−4
−6 29.5 λ
−8 a
−90 −60 −30 0 30 60 90
z
ϑ [deg] 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 30
x
a O

16 14.8
14.8 dBi
dBi 14.8
14.8 dBi
dBi 14.8
14.8 dBi
dBi 29 λ
ϑsc = 0◦
14
ϑsc = 30 ◦ b
12 z
ϑsc = 60◦
10 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 30
x
8 O
|D(ϑ)| [dBi]

6
4
29 λ
2
0 c
−2
−4
Fig. 6. Thinned arrays, with some of the elements being replaced by X3-
−6 sa-s. (a) DynP placement + 4 X3-sa-s; (b) CDS placement + 4 X3-sa-s; (c)
−8 CDS placement + 7 X3-sa-s.
−90 −60 −30 0 30 60 90
ϑ [deg]
The AF-s corresponding to the linear arrays in Fig. 6
b
are shown in Fig. 7, with the key array performance metrics
Fig. 5. Array factors (κ = 1) of the thinned arrays in Fig. 4. (a) DynP being summarised in Table II. As expected, D(ϑ)max slightly
placement; (b) CDS placement. diminishes at broadside but it is enhanced at 60◦ , with the gain
being larger in the case of adding 7 X3-sa-s (1.2 dB) when
compared to the case of adding only 4 X3-sa-s (0.7 dB). The
The AF-s of the two linear arrays are shown in Fig. 5. D(ϑ)max enhancement is remarkably similar when comparing
Furthermore, the following key array performance metrics the DynP and CDS arrays, which allows conjecturing that
are given in Table I: the maximum directivity D(ϑ)max , the incorporating X3-sa-s will yield similar results in conjunction
beamwidth (BW), the peak sidelobes level (peak SLL) and the with any sparse linear array that can accommodate them,
first sidelobes level (first SLL), namely the level of the lobes irrespective of the employed placement algorithm.
adjacent to the main beam (the peak SLL and first SLL being
normalised with respect to the main beam). TABLE II. K EY ARRAY PERFORMANCE METRICS FOR THE ARRAY
FACTORS IN F IG . 7.

As expected, the patterns are GL-free up to 60 beam
scanning. The obtained D(ϑ)max is the same, a direct conse- D(ϑ)max Beamwidth First SLL Peak SL
quence of both arrays having the same number of elements (see DynP placement + 4 X3-sa-s
[Section V] [12]). Moreover, D(ϑ)max does not change with ϑs = 0◦ 14.4 dBi 2.8◦ -16 dB -13.9 dB
scanning. The BW of the CDS array is consistently narrower, ϑs = 30◦ 14.8 dBi 3◦ -16.3 dB -13.6 dB

ϑs = 60 15.1 dBi 4.9◦ -16.4 dB -14.3 dB
this deviating from the observation in [Section III] [12] that
relates the BW to the array length (practically, equal in this CDS placement + 4 X3-sa-s
ϑs = 0◦ 14.4 dBi 1.9◦ -12.6 dB -12.1 dB
case). This situation is attributed to the elements being more ϑs = 30◦ 14.8 dBi 2.1◦ -13.4 dB -12.3 dB
evenly spread in the CDS configuration (thus, closer to the ϑs = 60◦ 15.1 dBi 3.7◦ -14 dB -13.1 dB
cases examined in [12]), while the DynP placement has most CDS placement + 7 X3-sa-s
elements concentrated at the arrays centre, with just some loose ϑs = 0◦ 14.1 dBi 1.9◦ -12.4 dB -11.4 dB
elements towards the edges. On an overall, the DynP placement ϑs = 30◦ 14.7 dBi 2.1◦ -13.4 dB -11.9 dB
performs better in terms of sidelobes level (SLL). ϑs = 60◦ 15.3 dBi 3.7◦ -14.1 dB -13.1 dB

978-1-5090-0783-7/16/$31.00 ©2016 IEEE


2016 Loughborough Antennas & Propagation Conference (LAPC)

14.4
14.4 dBi
dBi 14.8
14.8 dBi
dBi 15.1
15.1 dBi
dBi 24 22.2
22.2 dBi
dBi 21.3
21.3 dBi
dBi 16.8
16.8 dBi
dBi
16
ϑsc = 0◦ 22 ϑsc = 0◦
14 ◦ 20 ◦
ϑsc = 30 ϑsc = 30
12 18
ϑsc = 60◦ ϑsc = 60◦
10 16
14
8
12
|D(ϑ)| [dBi]

|D(ϑ)| [dBi]
6 10
4 8
2 6
4
0 2
−2 0
−4 −2
−4
−6
−6
−8 −8
−90 −60 −30 0 30 60 90 −90 −60 −30 0 30 60 90
ϑ [deg] ϑ [deg]

a a
14.4
14.4 dBi
dBi 14.8
14.8 dBi
dBi 15.1
15.1 dBi
dBi 24 22.2
22.2 dBi
dBi 21.3
21.3 dBi
dBi 16.8
16.8 dBi
dBi
16
ϑsc = 0◦ 22 ϑsc = 0◦
14 20
ϑsc = 30◦ ϑsc = 30◦
12 18
ϑsc = 60◦ ϑsc = 60◦
10 16
14
8
12
|D(ϑ)| [dBi]

|D(ϑ)| [dBi]
6 10
4 8
2 6
4
0 2
−2 0
−4 −2
−4
−6
−6
−8 −8
−90 −60 −30 0 30 60 90 −90 −60 −30 0 30 60 90
ϑ [deg] ϑ [deg]
b b
14.1
14.1 dBi
dBi 14.7
14.7 dBi
dBi 15.3
15.3 dBi
dBi 24 21.9
21.9 dBi
dBi 21.3
21.3 dBi
dBi 17.0
17.0 dBi
dBi
16
ϑsc = 0◦ 22 ϑsc = 0◦
14 ◦ 20 ◦
ϑsc = 30 ϑsc = 30
12 18
ϑsc = 60◦ ϑsc = 60◦
10 16
14
8
12
|D(ϑ)| [dBi]

|D(ϑ)| [dBi]
6 10
4 8
2 6
4
0 2
−2 0
−4 −2
−4
−6
−6
−8 −8
−90 −60 −30 0 30 60 90 −90 −60 −30 0 30 60 90
ϑ [deg] ϑ [deg]

c c
Fig. 7. Array factors (κ = 1) of the thinned arrays in Fig. 6. (a) Fig. 8. Radiation patterns (κ = 2) of the thinned arrays in Fig. 6 in the case
DynP placement + 4 X3-sa-s; (b) CDS placement+4 X3-sa-s; (c) CDS of using a cos(ϑ)-type elementary radiator. (a) DynP placement + 4 X3-sa-s;
placement+7 X3-sa-s. (b) CDS placement+4 X3-sa-s; (c) CDS placement+7 X3-sa-s.

TABLE III. K EY ARRAY PERFORMANCE METRICS FOR THE ARRAY


To conclude with, the use of X3-sa-s in the case of FACTORS IN F IG . 8.
cos(ϑ)-type elementary radiators is examining via the pattern-
D(ϑ)max Beamwidth First SLL Peak SL
multiplied array patterns shown in Fig. 8, the key array
performance metrics being summarised in Table III. The be- DynP placement + 4 X3-sa-s
haviour of D(ϑ)max evidences the -6 dB drop in the radiation ϑs = 0◦ 22.2 dBi 1.9◦ -16.1 dB -15.9 dB
ϑs = 30◦ 21.3 dBi 2.1◦ -15.9 dB -14.5 dB
pattern at 60◦ . Moreover, this drop also affects the peak ϑs = 60◦ 16.8 dBi 3.7◦ -14.2 dB -10.1 dB
SLL that becomes -10 dB, -7.8 dB and -8.7 dB for the cases CDS placement + 4 X3-sa-s
of DynP+4 X3-sa-s, CDS+4 X3-sa-s and CDS+7 X3-sa-s ϑs = 0◦ 22.2 dBi 1.9◦ -12.7 dB -12.7 dB
arrays, respectively. This metric is insufficient for radar ap- ϑs = 30◦ 21.3 dBi 2.1◦ -13 dB -12.7 dB
plications and its improvement is a subject of future research. ϑs = 60◦ 16.8 dBi 3.7◦ -12.7 dB -7.9 dB
CDS placement + 7 X3-sa-s
ϑs = 0◦ 21.9 dBi 1.9◦ -12.5 dB -12.5 dB
V. C ONCLUSIONS ϑs = 30◦ 21.3 dBi 2.1◦ -13 dB -13 dB
ϑs = 60◦ 17 dBi 3.7◦ -12.9 dB -8.8 dB
A novel approach for compensating the scan loss in wide-
angular scanning array antennas due to the pattern falloff of
physical radiators was presented. The design philosophy was
based on replacing elements in (linear) arrays with subarrays arrays. An optimised subarray, with 5 dB radiation intensity
offering an enhanced radiation intensity at large angles. The enhancement at 60◦ was designed. It was incorporated into two
space for accommodating the subarrays was provided by linear array designed by means of dynamic programming and
resorting to suitable thinning techniques yielding a 50% reduc- Cyclic Differences Sets placement methods. An up to 1.2 dB
tion of the number of elements when compared with uniform improvement of the maximum directivity at 60◦ beam scanning

978-1-5090-0783-7/16/$31.00 ©2016 IEEE


2016 Loughborough Antennas & Propagation Conference (LAPC)

was demonstrated. The results are opportune for improving the [5] R. N. Bracewell, The Fourier Transform and Its Applications, Boston:
performance of wide-angular scanning radar array antennas. McGraw-Hill, 2000.
[6] R. L. Haupt, “Thinned arrays using genetic algorithms,” “Optimized
weighting of uniform subarrays of unequal sizes,” IEEE Trans. Antennas
ACKNOWLEDGMENT Propag., vol. 42, no. 7, pp. 993–999, July 1994.
The research reported in this paper was carried out during [7] G. Oliveri, M. Donelli, and A. Massa, “Linear array thinning exploiting
a placement of Fannush Shofi Akbar at the Delft University of almost difference sets,” IEEE Trans. Antennas Propag., vol. 57, no. 12,
pp. 3800–3812, Dec. 2009.
Technology that was financed Directorate General of Higher
[8] O. M. Bucci, M. D’Urso, T. Isernia, P. Angeletti, and G. Toso, “Deter-
Education of Indonesia by the scheme of Sandwich-Like and ministic synthesis of uniform amplitude sparse arrays via new density
PMDSU Scholarship. The received financial support is greatly taper techniques,” IEEE Trans. Antennas Propag., vol. 58, no. 6,
acknowledged. pp. 1949–1958, June 2010.
[9] B. Fuchs, “Application of convex relaxation to array synthesis prob-
R EFERENCES lems,” IEEE Trans. Antennas Propag., vol. 62, no. 2, pp. 634–640,
Feb. 2014.
[1] M. I. Skolnik, G. Nemhauser and J. W. Sherman, III, “Dynamic pro- [10] M. Simeoni, J. H. Dickhof, and I. E. Lager, “Investigation of the poten-
gramming applied to unequally-spaced arrays,” IRE Trans. Antennas tial for reconfigurability of the non-uniform, linear array antennas,” in
Propagat., vol. AP-12, pp. 35–43, Jan. 1964. Proc. 29 th ESA Antenna Workshop on Multiple Beams and Reconfig-
[2] D. G. Leeper, “Isophoric arrays - massively thinned phased arrays with urable Antennas, Noordwijk, the Netherlands, Apr. 2007, pp. 227–230.
well-controlled sidelobes,” IEEE Trans. Antennas Propag., vol. 47, [11] D. Gordon, “La Jolla cyclic difference set repository,” [Online]. Avail-
no. 3, pp. 1825–1835, Dec. 1999. able: www.ccrwest.org/diffsets.html.
[3] I. E. Lager, C. Trampuz, M. Simeoni, and L. P. Ligthart, “Interleaved [12] C. I. Coman, I. E. Lager, and L. P. Ligthart, “Design considerations
array antennas for FMCW radar applications,” IEEE Trans. Antennas in sparse array antennas,” in Proc. 3 rd EuRAD, Manchester, UK,
Propag., vol. 57, no. 8, pp. 2486–2490, Aug. 2009. Sep. 2006, pp. 72–75.
[4] C. A. Balanis, Antenna Theory: Analysis and Design, 3rd ed., New
York: John Wiley & Sons Inc., 2005.

978-1-5090-0783-7/16/$31.00 ©2016 IEEE

You might also like